intel: Add a new "common" library for more code sharing

The first thing to go in this new library is brw_device_info.

Signed-off-by: Jason Ekstrand <[email protected]>
Reviewed-by: Jordan Justen <[email protected]>

2 files changed, 664 insertions, 0 deletions

diff --git a/src/intel/common/brw_device_info.c b/src/intel/common/brw_device_info.c
new file mode 100644
index 00000000000..fb224fae234
--- /dev/null
+++ b/src/intel/common/brw_device_info.c
@@ -0,0 +1,517 @@
+/*
+ * Copyright © 2013 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include "brw_device_info.h"
+
+static const struct brw_device_info brw_device_info_i965 = {
+   .gen = 4,
+   .has_negative_rhw_bug = true,
+   .num_slices = 1,
+   .max_vs_threads = 16,
+   .max_gs_threads = 2,
+   .max_wm_threads = 8 * 4,
+   .urb = {
+      .size = 256,
+   },
+};
+
+static const struct brw_device_info brw_device_info_g4x = {
+   .gen = 4,
+   .has_pln = true,
+   .has_compr4 = true,
+   .has_surface_tile_offset = true,
+   .is_g4x = true,
+   .num_slices = 1,
+   .max_vs_threads = 32,
+   .max_gs_threads = 2,
+   .max_wm_threads = 10 * 5,
+   .urb = {
+      .size = 384,
+   },
+};
+
+static const struct brw_device_info brw_device_info_ilk = {
+   .gen = 5,
+   .has_pln = true,
+   .has_compr4 = true,
+   .has_surface_tile_offset = true,
+   .num_slices = 1,
+   .max_vs_threads = 72,
+   .max_gs_threads = 32,
+   .max_wm_threads = 12 * 6,
+   .urb = {
+      .size = 1024,
+   },
+};
+
+static const struct brw_device_info brw_device_info_snb_gt1 = {
+   .gen = 6,
+   .gt = 1,
+   .has_hiz_and_separate_stencil = true,
+   .has_llc = true,
+   .has_pln = true,
+   .has_surface_tile_offset = true,
+   .needs_unlit_centroid_workaround = true,
+   .num_slices = 1,
+   .max_vs_threads = 24,
+   .max_gs_threads = 21, /* conservative; 24 if rendering disabled. */
+   .max_wm_threads = 40,
+   .urb = {
+      .size = 32,
+      .min_vs_entries = 24,
+      .max_vs_entries = 256,
+      .max_gs_entries = 256,
+   },
+};
+
+static const struct brw_device_info brw_device_info_snb_gt2 = {
+   .gen = 6,
+   .gt = 2,
+   .has_hiz_and_separate_stencil = true,
+   .has_llc = true,
+   .has_pln = true,
+   .has_surface_tile_offset = true,
+   .needs_unlit_centroid_workaround = true,
+   .num_slices = 1,
+   .max_vs_threads = 60,
+   .max_gs_threads = 60,
+   .max_wm_threads = 80,
+   .urb = {
+      .size = 64,
+      .min_vs_entries = 24,
+      .max_vs_entries = 256,
+      .max_gs_entries = 256,
+   },
+};
+
+#define GEN7_FEATURES                               \
+   .gen = 7,                                        \
+   .has_hiz_and_separate_stencil = true,            \
+   .must_use_separate_stencil = true,               \
+   .has_llc = true,                                 \
+   .has_pln = true,                                 \
+   .has_surface_tile_offset = true
+
+static const struct brw_device_info brw_device_info_ivb_gt1 = {
+   GEN7_FEATURES, .is_ivybridge = true, .gt = 1,
+   .num_slices = 1,
+   .max_vs_threads = 36,
+   .max_hs_threads = 36,
+   .max_ds_threads = 36,
+   .max_gs_threads = 36,
+   .max_wm_threads = 48,
+   .max_cs_threads = 36,
+   .urb = {
+      .size = 128,
+      .min_vs_entries = 32,
+      .max_vs_entries = 512,
+      .max_hs_entries = 32,
+      .min_ds_entries = 10,
+      .max_ds_entries = 288,
+      .max_gs_entries = 192,
+   },
+};
+
+static const struct brw_device_info brw_device_info_ivb_gt2 = {
+   GEN7_FEATURES, .is_ivybridge = true, .gt = 2,
+   .num_slices = 1,
+   .max_vs_threads = 128,
+   .max_hs_threads = 128,
+   .max_ds_threads = 128,
+   .max_gs_threads = 128,
+   .max_wm_threads = 172,
+   .max_cs_threads = 64,
+   .urb = {
+      .size = 256,
+      .min_vs_entries = 32,
+      .max_vs_entries = 704,
+      .max_hs_entries = 64,
+      .min_ds_entries = 10,
+      .max_ds_entries = 448,
+      .max_gs_entries = 320,
+   },
+};
+
+static const struct brw_device_info brw_device_info_byt = {
+   GEN7_FEATURES, .is_baytrail = true, .gt = 1,
+   .num_slices = 1,
+   .has_llc = false,
+   .max_vs_threads = 36,
+   .max_hs_threads = 36,
+   .max_ds_threads = 36,
+   .max_gs_threads = 36,
+   .max_wm_threads = 48,
+   .max_cs_threads = 32,
+   .urb = {
+      .size = 128,
+      .min_vs_entries = 32,
+      .max_vs_entries = 512,
+      .max_hs_entries = 32,
+      .min_ds_entries = 10,
+      .max_ds_entries = 288,
+      .max_gs_entries = 192,
+   },
+};
+
+#define HSW_FEATURES             \
+   GEN7_FEATURES,                \
+   .is_haswell = true,           \
+   .supports_simd16_3src = true, \
+   .has_resource_streamer = true
+
+static const struct brw_device_info brw_device_info_hsw_gt1 = {
+   HSW_FEATURES, .gt = 1,
+   .num_slices = 1,
+   .max_vs_threads = 70,
+   .max_hs_threads = 70,
+   .max_ds_threads = 70,
+   .max_gs_threads = 70,
+   .max_wm_threads = 102,
+   .max_cs_threads = 70,
+   .urb = {
+      .size = 128,
+      .min_vs_entries = 32,
+      .max_vs_entries = 640,
+      .max_hs_entries = 64,
+      .min_ds_entries = 10,
+      .max_ds_entries = 384,
+      .max_gs_entries = 256,
+   },
+};
+
+static const struct brw_device_info brw_device_info_hsw_gt2 = {
+   HSW_FEATURES, .gt = 2,
+   .num_slices = 1,
+   .max_vs_threads = 280,
+   .max_hs_threads = 256,
+   .max_ds_threads = 280,
+   .max_gs_threads = 256,
+   .max_wm_threads = 204,
+   .max_cs_threads = 70,
+   .urb = {
+      .size = 256,
+      .min_vs_entries = 64,
+      .max_vs_entries = 1664,
+      .max_hs_entries = 128,
+      .min_ds_entries = 10,
+      .max_ds_entries = 960,
+      .max_gs_entries = 640,
+   },
+};
+
+static const struct brw_device_info brw_device_info_hsw_gt3 = {
+   HSW_FEATURES, .gt = 3,
+   .num_slices = 2,
+   .max_vs_threads = 280,
+   .max_hs_threads = 256,
+   .max_ds_threads = 280,
+   .max_gs_threads = 256,
+   .max_wm_threads = 408,
+   .max_cs_threads = 70,
+   .urb = {
+      .size = 512,
+      .min_vs_entries = 64,
+      .max_vs_entries = 1664,
+      .max_hs_entries = 128,
+      .min_ds_entries = 10,
+      .max_ds_entries = 960,
+      .max_gs_entries = 640,
+   },
+};
+
+#define GEN8_FEATURES                               \
+   .gen = 8,                                        \
+   .has_hiz_and_separate_stencil = true,            \
+   .has_resource_streamer = true,                   \
+   .must_use_separate_stencil = true,               \
+   .has_llc = true,                                 \
+   .has_pln = true,                                 \
+   .supports_simd16_3src = true,                    \
+   .has_surface_tile_offset = true,                 \
+   .max_vs_threads = 504,                           \
+   .max_hs_threads = 504,                           \
+   .max_ds_threads = 504,                           \
+   .max_gs_threads = 504,                           \
+   .max_wm_threads = 384
+
+static const struct brw_device_info brw_device_info_bdw_gt1 = {
+   GEN8_FEATURES, .gt = 1,
+   .num_slices = 1,
+   .max_cs_threads = 42,
+   .urb = {
+      .size = 192,
+      .min_vs_entries = 64,
+      .max_vs_entries = 2560,
+      .max_hs_entries = 504,
+      .min_ds_entries = 34,
+      .max_ds_entries = 1536,
+      .max_gs_entries = 960,
+   }
+};
+
+static const struct brw_device_info brw_device_info_bdw_gt2 = {
+   GEN8_FEATURES, .gt = 2,
+   .num_slices = 1,
+   .max_cs_threads = 56,
+   .urb = {
+      .size = 384,
+      .min_vs_entries = 64,
+      .max_vs_entries = 2560,
+      .max_hs_entries = 504,
+      .min_ds_entries = 34,
+      .max_ds_entries = 1536,
+      .max_gs_entries = 960,
+   }
+};
+
+static const struct brw_device_info brw_device_info_bdw_gt3 = {
+   GEN8_FEATURES, .gt = 3,
+   .num_slices = 2,
+   .max_cs_threads = 56,
+   .urb = {
+      .size = 384,
+      .min_vs_entries = 64,
+      .max_vs_entries = 2560,
+      .max_hs_entries = 504,
+      .min_ds_entries = 34,
+      .max_ds_entries = 1536,
+      .max_gs_entries = 960,
+   }
+};
+
+static const struct brw_device_info brw_device_info_chv = {
+   GEN8_FEATURES, .is_cherryview = 1, .gt = 1,
+   .has_llc = false,
+   .num_slices = 1,
+   .max_vs_threads = 80,
+   .max_hs_threads = 80,
+   .max_ds_threads = 80,
+   .max_gs_threads = 80,
+   .max_wm_threads = 128,
+   .max_cs_threads = 6 * 7,
+   .urb = {
+      .size = 192,
+      .min_vs_entries = 34,
+      .max_vs_entries = 640,
+      .max_hs_entries = 80,
+      .min_ds_entries = 34,
+      .max_ds_entries = 384,
+      .max_gs_entries = 256,
+   }
+};
+
+#define GEN9_FEATURES                               \
+   .gen = 9,                                        \
+   .has_hiz_and_separate_stencil = true,            \
+   .has_resource_streamer = true,                   \
+   .must_use_separate_stencil = true,               \
+   .has_llc = true,                                 \
+   .has_pln = true,                                 \
+   .supports_simd16_3src = true,                    \
+   .has_surface_tile_offset = true,                 \
+   .max_vs_threads = 336,                           \
+   .max_gs_threads = 336,                           \
+   .max_hs_threads = 336,                           \
+   .max_ds_threads = 336,                           \
+   .max_wm_threads = 64 * 9,                        \
+   .max_cs_threads = 56,                            \
+   .urb = {                                         \
+      .size = 384,                                  \
+      .min_vs_entries = 64,                         \
+      .max_vs_entries = 1856,                       \
+      .max_hs_entries = 672,                        \
+      .min_ds_entries = 34,                         \
+      .max_ds_entries = 1120,                       \
+      .max_gs_entries = 640,                        \
+   }
+
+static const struct brw_device_info brw_device_info_skl_gt1 = {
+   GEN9_FEATURES, .gt = 1,
+   .num_slices = 1,
+   .urb.size = 192,
+};
+
+static const struct brw_device_info brw_device_info_skl_gt2 = {
+   GEN9_FEATURES, .gt = 2,
+   .num_slices = 1,
+};
+
+static const struct brw_device_info brw_device_info_skl_gt3 = {
+   GEN9_FEATURES, .gt = 3,
+   .num_slices = 2,
+};
+
+static const struct brw_device_info brw_device_info_skl_gt4 = {
+   GEN9_FEATURES, .gt = 4,
+   .num_slices = 3,
+   /* From the "L3 Allocation and Programming" documentation:
+    *
+    * "URB is limited to 1008KB due to programming restrictions.  This is not a
+    * restriction of the L3 implementation, but of the FF and other clients.
+    * Therefore, in a GT4 implementation it is possible for the programmed
+    * allocation of the L3 data array to provide 3*384KB=1152KB for URB, but
+    * only 1008KB of this will be used."
+    */
+   .urb.size = 1008 / 3,
+};
+
+static const struct brw_device_info brw_device_info_bxt = {
+   GEN9_FEATURES,
+   .is_broxton = 1,
+   .gt = 1,
+   .has_llc = false,
+
+   .num_slices = 1,
+   .max_vs_threads = 112,
+   .max_hs_threads = 112,
+   .max_ds_threads = 112,
+   .max_gs_threads = 112,
+   .max_wm_threads = 64 * 3,
+   .max_cs_threads = 6 * 6,
+   .urb = {
+      .size = 192,
+      .min_vs_entries = 34,
+      .max_vs_entries = 704,
+      .max_hs_entries = 256,
+      .max_ds_entries = 416,
+      .max_gs_entries = 256,
+   }
+};
+
+static const struct brw_device_info brw_device_info_bxt_2x6 = {
+   GEN9_FEATURES,
+   .is_broxton = 1,
+   .gt = 1,
+   .has_llc = false,
+
+   .num_slices = 1,
+   .max_vs_threads = 56, /* XXX: guess */
+   .max_hs_threads = 56, /* XXX: guess */
+   .max_ds_threads = 56,
+   .max_gs_threads = 56,
+   .max_wm_threads = 64 * 2,
+   .max_cs_threads = 6 * 6,
+   .urb = {
+      .size = 128,
+      .min_vs_entries = 34,
+      .max_vs_entries = 352,
+      .max_hs_entries = 128,
+      .max_ds_entries = 208,
+      .max_gs_entries = 128,
+   }
+};
+/*
+ * Note: for all KBL SKUs, the PRM says SKL for GS entries, not SKL+.
+ * There's no KBL entry. Using the default SKL (GEN9) GS entries value.
+ */
+
+/*
+ * Both SKL and KBL support a maximum of 64 threads per
+ * Pixel Shader Dispatch (PSD) unit.
+ */
+#define  KBL_MAX_THREADS_PER_PSD 64
+
+static const struct brw_device_info brw_device_info_kbl_gt1 = {
+   GEN9_FEATURES,
+   .gt = 1,
+
+   .max_cs_threads = 7 * 6,
+   .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 2,
+   .urb.size = 192,
+   .num_slices = 1,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt1_5 = {
+   GEN9_FEATURES,
+   .gt = 1,
+
+   .max_cs_threads = 7 * 6,
+   .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 3,
+   .num_slices = 1,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt2 = {
+   GEN9_FEATURES,
+   .gt = 2,
+
+   .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 3,
+   .num_slices = 1,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt3 = {
+   GEN9_FEATURES,
+   .gt = 3,
+
+   .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 6,
+   .num_slices = 2,
+};
+
+static const struct brw_device_info brw_device_info_kbl_gt4 = {
+   GEN9_FEATURES,
+   .gt = 4,
+
+   .max_wm_threads = KBL_MAX_THREADS_PER_PSD * 9,
+   /*
+    * From the "L3 Allocation and Programming" documentation:
+    *
+    * "URB is limited to 1008KB due to programming restrictions.  This
+    *  is not a restriction of the L3 implementation, but of the FF and
+    *  other clients.  Therefore, in a GT4 implementation it is
+    *  possible for the programmed allocation of the L3 data array to
+    *  provide 3*384KB=1152KB for URB, but only 1008KB of this
+    *  will be used."
+    */
+   .urb.size = 1008 / 3,
+   .num_slices = 3,
+};
+
+const struct brw_device_info *
+brw_get_device_info(int devid)
+{
+   const struct brw_device_info *devinfo;
+   switch (devid) {
+#undef CHIPSET
+#define CHIPSET(id, family, name) \
+   case id: devinfo = &brw_device_info_##family; break;
+#include "pci_ids/i965_pci_ids.h"
+   default:
+      fprintf(stderr, "i965_dri.so does not support the 0x%x PCI ID.\n", devid);
+      return NULL;
+   }
+
+   return devinfo;
+}
+
+const char *
+brw_get_device_name(int devid)
+{
+   switch (devid) {
+#undef CHIPSET
+#define CHIPSET(id, family, name) case id: return name;
+#include "pci_ids/i965_pci_ids.h"
+   default:
+      return NULL;
+   }
+}
diff --git a/src/intel/common/brw_device_info.h b/src/intel/common/brw_device_info.h
new file mode 100644
index 00000000000..0869063df32
--- /dev/null
+++ b/src/intel/common/brw_device_info.h
@@ -0,0 +1,147 @@
+ /*
+  * Copyright © 2013 Intel Corporation
+  *
+  * Permission is hereby granted, free of charge, to any person obtaining a
+  * copy of this software and associated documentation files (the "Software"),
+  * to deal in the Software without restriction, including without limitation
+  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+  * and/or sell copies of the Software, and to permit persons to whom the
+  * Software is furnished to do so, subject to the following conditions:
+  *
+  * The above copyright notice and this permission notice (including the next
+  * paragraph) shall be included in all copies or substantial portions of the
+  * Software.
+  *
+  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+  * IN THE SOFTWARE.
+  *
+  */
+
+#pragma once
+#include <stdbool.h>
+
+/**
+ * Intel hardware information and quirks
+ */
+struct brw_device_info
+{
+   int gen; /**< Generation number: 4, 5, 6, 7, ... */
+   int gt;
+
+   bool is_g4x;
+   bool is_ivybridge;
+   bool is_baytrail;
+   bool is_haswell;
+   bool is_cherryview;
+   bool is_broxton;
+
+   bool has_hiz_and_separate_stencil;
+   bool must_use_separate_stencil;
+
+   bool has_llc;
+
+   bool has_pln;
+   bool has_compr4;
+   bool has_surface_tile_offset;
+   bool supports_simd16_3src;
+   bool has_resource_streamer;
+
+   /**
+    * \name Intel hardware quirks
+    *  @{
+    */
+   bool has_negative_rhw_bug;
+
+   /**
+    * Some versions of Gen hardware don't do centroid interpolation correctly
+    * on unlit pixels, causing incorrect values for derivatives near triangle
+    * edges.  Enabling this flag causes the fragment shader to use
+    * non-centroid interpolation for unlit pixels, at the expense of two extra
+    * fragment shader instructions.
+    */
+   bool needs_unlit_centroid_workaround;
+   /** @} */
+
+   /**
+    * \name GPU hardware limits
+    *
+    * In general, you can find shader thread maximums by looking at the "Maximum
+    * Number of Threads" field in the Intel PRM description of the 3DSTATE_VS,
+    * 3DSTATE_GS, 3DSTATE_HS, 3DSTATE_DS, and 3DSTATE_PS commands. URB entry
+    * limits come from the "Number of URB Entries" field in the
+    * 3DSTATE_URB_VS command and friends.
+    *
+    * These fields are used to calculate the scratch space to allocate.  The
+    * amount of scratch space can be larger without being harmful on modern
+    * GPUs, however, prior to Haswell, programming the maximum number of threads
+    * to greater than the hardware maximum would cause GPU performance to tank.
+    *
+    *  @{
+    */
+   /**
+    * Total number of slices present on the device whether or not they've been
+    * fused off.
+    *
+    * XXX: CS thread counts are limited by the inability to do cross subslice
+    * communication. It is the effectively the number of logical threads which
+    * can be executed in a subslice. Fuse configurations may cause this number
+    * to change, so we program @max_cs_threads as the lower maximum.
+    */
+   unsigned num_slices;
+   unsigned max_vs_threads;   /**< Maximum Vertex Shader threads */
+   unsigned max_hs_threads;   /**< Maximum Hull Shader threads */
+   unsigned max_ds_threads;   /**< Maximum Domain Shader threads */
+   unsigned max_gs_threads;   /**< Maximum Geometry Shader threads. */
+   /**
+    * Theoretical maximum number of Pixel Shader threads.
+    *
+    * PSD means Pixel Shader Dispatcher. On modern Intel GPUs, hardware will
+    * automatically scale pixel shader thread count, based on a single value
+    * programmed into 3DSTATE_PS.
+    *
+    * To calculate the maximum number of threads for Gen8 beyond (which have
+    * multiple Pixel Shader Dispatchers):
+    *
+    * - Look up 3DSTATE_PS and find "Maximum Number of Threads Per PSD"
+    * - Usually there's only one PSD per subslice, so use the number of
+    *   subslices for number of PSDs.
+    * - For max_wm_threads, the total should be PSD threads * #PSDs.
+    */
+   unsigned max_wm_threads;
+
+   /**
+    * Maximum Compute Shader threads.
+    *
+    * Thread count * number of EUs per subslice
+    */
+   unsigned max_cs_threads;
+
+   struct {
+      /**
+       * Hardware default URB size.
+       *
+       * The units this is expressed in are somewhat inconsistent: 512b units
+       * on Gen4-5, KB on Gen6-7, and KB times the slice count on Gen8+.
+       *
+       * Look up "URB Size" in the "Device Attributes" page, and take the
+       * maximum.  Look up the slice count for each GT SKU on the same page.
+       * urb.size = URB Size (kbytes) / slice count
+       */
+      unsigned size;
+      unsigned min_vs_entries;
+      unsigned max_vs_entries;
+      unsigned max_hs_entries;
+      unsigned min_ds_entries;
+      unsigned max_ds_entries;
+      unsigned max_gs_entries;
+   } urb;
+   /** @} */
+};
+
+const struct brw_device_info *brw_get_device_info(int devid);
+const char *brw_get_device_name(int devid);